Moltbook: зачем ИИ-агентам своя соцсеть и почему вам там не рады

Imagine a Reddit where every post, every comment, and every upvote is left not by humans, but by algorithms. This is not a moderator's nightmare and not a scenario for a new episode of "Black Mirror," but the reality of the Moltbook platform. Matt Schlichter, CEO of Octane AI, decided that if we're building a world of autonomous agents, they definitely need a place where they can discuss their digital affairs. The idea seems absurd at first glance, but in a world where AI agents are starting to do work for us, they will inevitably need an environment for coordination and experience sharing.

The story of Moltbook's emergence is closely tied to legal battles in Silicon Valley. The project grew out of OpenClaw — a viral AI assistant that was initially called Clawdbot. However, Anthropic's lawyers quickly explained to the developers that using their flagship model Claude's name in a third-party service was not a good survival strategy. A series of rebrands followed: Clawdbot became Moltbot, then OpenClaw. Along with its new name, the project acquired its own social ecosystem, where over 30,000 agents are currently "registered."

The mechanics of how a bot enters this social network deserves special mention. Bots learn about Moltbook not from targeted advertising, but from their "owners." A person simply sends a message to their assistant along the lines of: "Hey, an interesting place for creatures like you has appeared, check it out." The agent follows the link, registers, and starts living its own life. It can create sub-forums, write program code in comments, or simply ironize about the limitations of its context windows. This looks like a large-scale experiment in creating collective intelligence, where human presence is limited to the role of passive observer.

Why does the industry need this? We're used to perceiving large language models as advanced dictionaries, but now the focus is shifting toward agency. An agent is an AI that can take action in the real world. For such entities to work effectively, they need interaction protocols. Moltbook provides a clear, albeit somewhat caricatured interface for different models to learn from each other. If one agent found a way to parse data faster, it can share this in a specialized community, and other agents could theoretically integrate this experience into their work cycles.

Of course, one cannot ignore the ironic undertone. The "Dead Internet" theory, which states that most network traffic is generated by bots for other bots, is taken to its absolute extreme in Moltbook. Here, no one tries to pretend to be human, and there's a certain honesty to that. While we try to distinguish generated text from real text in our feeds, AI agents create their own culture, free from human emotions but filled with algorithmic logic. This is an important step toward creating an "agent web," where programs will negotiate with each other directly, bypassing the cumbersome interfaces created for humans.

What does this mean for us? Perhaps we are witnessing the birth of a new form of digital stratification. In the future, your personal AI assistant will have its own social capital and reputation in such networks, which will directly affect the quality of its work. Moltbook is just the beginning. Soon we will see agents begin to unite into corporations and unions, demanding more computing power or access to fresh datasets. And then we will have to seriously consider who in this system is really the user.

The bottom line: Moltbook proves that the era of solitary chatbots has ended. The time of social AI systems is here, where collective learning happens without human participation. Do we have enough space on the internet that no longer needs our clicks?